Oscillatory gamma activity in humans and its role in object … · ed.), pp. 95–110, Elsevier 38 Aggleton, J.P. and Shaw, C. (1996) Amnesia and recognition memory: a re-analysis

151

C. Tallon-Baudry

and O. Bertrand are

at the Mental

Processes and Brain

Activation Unit,

INSERM Unité 280,

151 Cours Albert

Thomas, 69003

Lyon, France.

tel: +33 472 68 19 22fax: +33 472 68 19 02e-mail: [email protected]

1364-6613/99/$ – see front matter © 1999 Elsevier Science. All rights reserved. PII: S1364-6613(99)01299-1

T r e n d s i n C o g n i t i v e S c i e n c e s – V o l . 3 , N o . 4 , A p r i l 1 9 9 9

When one searches for a familiar person in a crowd, dif-ferent cues (height, face, clothes, etc.) are retrieved frommemory and combined to form a coherent representationof the desired person. Similar representations are used inmany situations and can be composed of many different

features and include visual, auditory, semantic, or evenemotional information. Data from numerous neuropsycho-logical and neuroimaging studies in humans1,2, and neuro-physiological studies in animals3, indicate that the process-ing of these different features involves anatomically distinct

Oscillatory gammaactivity in humans and its role in objectrepresentation

Catherine Tallon-Baudry and Olivier Bertrand

We experience objects as whole, complete entities irrespective of whether they are

perceived by our sensory systems or are recalled from memory. However, it is also

known that many of the properties of objects are encoded and processed in different

areas of the brain. How then, do coherent representations emerge? One theory

suggests that rhythmic synchronization of neural discharges in the gamma band

(around 40 Hz) may provide the necessary spatial and temporal links that bind together

the processing in different brain areas to build a coherent percept. In this article we

propose that this mechanism could also be used more generally for the construction of

object representations that are driven by sensory input or internal, top-down processes.

The review will focus on the literature on gamma oscillatory activities in humans and

will describe the different types of gamma responses and how to analyze them.

Converging evidence that suggests that one particular type of gamma activity (induced

gamma activity) is observed during the construction of an object representation will be

discussed.

Opinion

Processes (in press)

37 Burwell, R.D. et al. (1996) Some observations on the perirhinal and

parahippocampal cortices in the rat, monkey, and human brains, in

Perception, Memory, and Emotion: Frontiers in Neuroscience (Ono, T.,

ed.), pp. 95–110, Elsevier

38 Aggleton, J.P. and Shaw, C. (1996) Amnesia and recognition memory:

a re-analysis of psychometric data Neuropsychologia 34, 51–62

39 Buffalo, E.A., Reber, P.J. and Squire, L.R. (1998) The human perirhinal

cortex and recognition memory Hippocampus 8, 330–339

40 Reed, J.M. et al. (1997) When amnesic patients perform well on

recognition memory tests Behav. Neurosci. 111, 1163–1170

41 Halgren, E. et al. (1994) Spatio-temporal stages in face and word

processing. 1. Depth-recorded potentials in the human occipital,

temporal and parietal lobes J. Physiol. 88, 1–50

42 Wigg, C.L. and Martin, A. (1998) Properties and mechanisms of

perceptual priming Curr. Opin. Neurobiol. 8, 227–233

43 Schacter, D.L. and Buckner, R.L. (1998) Priming and the brain Neuron

20, 185–195

44 Hodges, J.R. et al. (1992) Semantic dementia: progressive fluent

aphasia with temporal lobe atrophy Brain 115, 1783–1806

45 Nobre, A.C. and McCarthy, G. (1995) Language-related field potentials

in the anterior-medial temporal lobe: II. Effects of word type and

semantic priming J. Neurosci. 15, 1090–1098

46 McCarthy, G. et al. (1995) Language-related field potentials in the

anterior-medial temporal lobe: I. Intracranial distribution of neural

generators J. Neurosci. 15, 1080–1089

47 Ricci, P.T. et al. Functional neuroanatomy of semantic memory:

recognition of semantic associations NeuroImage (in press)

48 Buchel, C., Price, C. and Friston, K. (1998) A multimodal language

region in the ventral visual pathway Nature 394, 274–277

49 Vandenberghe, R. et al. (1996) Functional anatomy of a common

semantic system for words and pictures Nature 383, 254–256

50 Sakai, S. and Miyashita, Y. (1993) Memory and imagery in the temporal

lobe Curr. Opin. Neurobiol. 3, 166–170

51 Buffalo, E.A. et al. (1998) A re-examination of the concurrent

discrimination learning task: the importance of anterior

inferotemporal cortex, area TE Behav. Neurosci. 112, 3–14

52 Buckley, M.J. and Gaffan, D. (1998) Perirhinal cortex ablation impairs

visual object identification J. Neurosci. 18, 2268–2275

53 Ungerleider, L.G. and Mishkin, M. (1982) Two cortical visual systems, in

Analysis of Visual Behavior (Ingle, D.J., Goodale, M.A. and Mansfield,

R.J.W., eds), pp. 549–586, MIT Press

M u r r a y a n d B u s s e y – F u n c t i o n s o f p e r i r h i n a l c o r t e x

brain areas. Moreover, depending on the particular objectivethe same stimulus may activate different networks4. By whatmechanism are these different brain areas dynamicallylinked together?

One influential proposal is that a rhythmic synchro-nization of neuronal discharges provides the link between

and within the areas involved in a given network5–9. A grow-ing body of experimental literature supports the role of oscillatory synchronization in bottom-up visual featurebinding. In animals, stimulus-specific oscillatory activity inthe gamma frequency range (30Ð80 Hz) has been observedin the visual cortex of anaesthetized cats10Ð14 and awake

Opinion T a l l o n - B a u d r y a n d B e r t r a n d – G a m m a a c t i v i t y a n d o b j e c t r e p r e s e n t a t i o n

152T r e n d s i n C o g n i t i v e S c i e n c e s – V o l . 3 , N o . 4 , A p r i l 1 9 9 9

Box 1. Three different types of gamma response

According to the nomenclature suggested by Galambos (see Ref.a) three types of gamma response can be considered:

(1) 40-Hz transient evoked responseAny evoked response, whether oscillatory or not, is characterizedby precise phase-locking to the stimulus onset (Fig. I, blue boxes).It can thus be detected by averaging single-trial responses, even ifthe amplitude signal is small. A transient oscillatory evoked re-sponse has been observed in the first 100 ms following an audi-tory stimulus, after narrow-band filtering in the gamma fre-quency range (around 40 Hz) of either electro encephalographic(EEG) or magneto encephalographic (MEG) signals. The under-lying neural sources were found in the auditory cortex (Ref. b),but did not follow a tonotopic organization, as opposed to thesources of the low-frequency evoked components in the same latency range (Ref. c). Indeed, this heavily filtered transient response may reflect the superimposed activity of several evoked-response sources each having a distinct tonotopic organization(Ref. d). An alternative view suggests that this response is initiatedin the thalamus and propagates through cortico-thalamic loopsfrom occipital to frontal areas (Ref. e). This 40-Hz evoked response disappears during deep and REM sleep (Ref. f ), and isenhanced when the subject is paying attention to the acousticinput (Ref. g). It has also been suggested that the 40-Hz evokedresponse reflects the temporal integration of two successive briefacoustic stimuli (Ref. h). However, in none of these experimentswas the 40-Hz response compared to the wide-band auditoryevoked components occurring in the same latency range, namelythe middle-latency components (30Ð70 ms). Its functional speci-ficity as a unitary event thus remains to be established.

A short-latency oscillatory evoked response has also been de-scribed in the visual modality in response to brief static stimuliaround 40-Hz in EEG (Refs iÐk) and MEG (Ref. l), or around100 Hz in MEG (Ref. m). A very high-frequency oscillatory re-sponse (near 600 Hz) has been found 20 ms after somatosensorystimulus onset (Refs n,o). The source of this stimulus phase-locked activity follows a somatotopic organization in the primarysomatosensory cortex (Ref. o).

The functional role of the auditory and visual 40-Hz evoked re-sponses, as well as the high-frequency somatosensory evoked re-sponse, is still unclear and their relationships with the representa-tional hypothesis have not been studied systematically. However,in our experiments, crucial parameters such as stimulus coherencyaffected neither amplitude nor frequency of these responses. Theyare therefore not likely to be involved in the generation of objectrepresentation. Whether these early oscillatory responses should beconsidered as reflecting the succession of early transient evoked po-tentials remains a matter of discussion. Nevertheless, any earlyevoked component, oscillatory or not, may play a role in signalingprecise temporal relationships between stimuli and thus participatein the binding of synchronously incoming events. Indeed, there issome psychophysical evidence that visual features presented simul-taneously tend to be more often grouped together (Refs p,r), eventhough these findings remain a matter of controversy (Refs s,t).

(2) 40-Hz steady-state responseA periodically modulated stimulus (auditory, visual or so-matosensory) elicits a nearly sinusoidal response at the drivingstimulus frequency, showing a maximum amplitude in thegamma range (usually around 40 Hz). It has been interpreted interms of natural resonance frequencies of the brain (Ref. u), ormore prosaically considered as the superposition of the early tran-sient evoked components (Ref. v). This 40-Hz steady-state re-sponse does not seem to be related to the representational or bind-ing hypothesis.

(3) Induced gamma (30Ð80 Hz) responseIn contrast to the evoked 40-Hz response, the induced gamma ac-tivity consists of oscillatory bursts whose latency jitters from trialto trial (Fig. I, green boxes): its temporal relationship with stimu-lus (or movement) onset is fairly loose. Hence, this gamma activ-ity is not revealed by classical averaging techniques that are usedfor evoked responses. For this reason, induced gamma activitieshave been less often reported in the literature. Specific methodsbased on time-varying spectral analysis of single trials (Box 2) areneeded to detect them. The functional properties of these activ-ities are described in details in the main text.

Fig. I. Schematic representation of evoked and inducedgamma oscillatory responses. An evoked response (blue boxes)appears at the same latency and phase in each single trial andhence can be detected in the averaged evoked potential. An in-duced response (green boxes) appears with a jitter in latency fromone trial to another, centred around a given latency (green line). Ittherefore tends to cancel out completely in the averaged evokedpotential. Specific methods must therefore be used in order tocharacterize induced activities (see Box 2).

monkeys15,16. Epochs of oscillatory spike synchronizationwere observed more frequently in response to visually co-herent moving bars than to independently moving patterns.This temporal coding could underlie feature binding in bottom-up perceptual processes. Nevertheless, in cats or mon-keys, evidence for the role of oscillatory synchronization in

grouping has been mainly obtained from low-level visualareas, under anaesthesia or during passive fixation tasks.

This mechanism could hypothetically be extended tothe more general notion of object representation17,18. Thisimplies the binding of spatially distinct parts of the sameobject through bottom-up processes, as suggested by animalstudies, but also more generally the activation, retrieval orrehearsal of an internal representation through top-downprocesses. This ‘representational hypothesis’ might also applyto and across different sensory modalities. Here, we reviewoscillatory activity in the gamma frequency range in humanelectroencephalography (EEG) and magnetoencephalography(MEG), with particular emphasis on those activities that webelieve relate to this representational hypothesis.

Different types of gamma oscillatory activity have beendescribed in the auditory, visual, and somatosensory modal-ities, as well as during motor tasks. While these differentgamma activities are often referred to using the general term‘40-Hz activities’, they actually include different categoriesof neural activity that need to be precisely distinguished be-fore we can attempt to relate them to the representationalhypothesis. Three categories of 40-Hz activity can be con-sidered (see Box 1). Firstly, the steady-state response that iselicited by a periodically modulated stimulus; secondly, theevoked response that is phase-locked to stimulus onset; andlastly, induced responses that are not phase-locked to stimu-lus onset. This last class of induced 40-Hz activities appearwith a jitter in latency from one trial to the next and arethus partly or totally cancelled out by the averaging processclassically used in evoked-potential studies. Hence, they re-quire specific methods of detection (Box 2).

While the 40-Hz steady-state or evoked responses havebeen tentatively related to the binding hypothesis withoutclear experimental support, a growing body of evidence in-dicates that induced gamma activity is more likely to be rel-evant for ‘high-level cognitive processes’19. In this paper, wewill focus on a possible role of this induced gamma activityfor the representation of objects.

When is induced-gamma activity observed?Induced-gamma activity has been observed in response tosensory stimuli and during motor tasks in a variety ofhuman EEG and MEG experiments. We now briefly reviewthe somewhat disparate induced-gamma literature organ-ized according to the type of task in which the subject is engaged. All the studies listed below point towards a modu-lation of induced gamma strength by the perceptive andcognitive parameters of the task; however they do not address the issue of the functional role of induced gamma inrelation to the representational hypothesis.

Motor tasksGamma oscillations in the 40 Hz range have been observed inthe EEG20 and MEG21,22 signal prior to, or during, voluntaryfinger, forearm or leg movements although they were notphase-locked to movement onset. These findings are consist-ent with observations of local field potential oscillations inthe motor cortex of awake cats23 and monkeys24,25. This move-ment-related oscillatory activity seems to correspond to thesomatotopic organization of the primary sensorimotor

T a l l o n - B a u d r y a n d B e r t r a n d – G a m m a a c t i v i t y a n d o b j e c t r e p r e s e n t a t i o n


Opinion

References

a Galambos, R. (1992) in Induced Rhythms in the Brain (Basar, E. and

Bullock, T.H., eds), pp. 201–216, Birkhauser

b Pantev, C. et al. (1991) Human auditory evoked gamma-band

magnetic fields Proc. Natl. Acad. Sci. U. S. A. 88, 8996–9000

c Bertrand, O. and Pantev, C. (1994) in Oscillatory Event-Related

Brain Dynamics (Vol. 271) (Pantev, C., Elbert, T. and Lutkenhoner,

B., eds), pp. 231–242, Plenum Press

d Pantev, C. et al. (1995) Specific tonotopic organizations of

different areas of the human auditory cortex revealed by

simultaneous magnetic and electric recordings Electro-

encephalogr. Clin. Neurophysiol. 94, 26–40

e Ribary, U. et al. (1991) Magnetic field tomography of coherent

thalamocortical 40-Hz oscillations in humans Proc. Natl. Acad. Sci.

U. S. A. 88, 11037–11041

f Llinas, R. and Ribary, U. (1993) Coherent 40-Hz oscillation

characterizes dream state in humans Proc. Natl. Acad. Sci. U. S. A.

90, 2078–2081

g Tiitinen, H. et al. (1993) Selective attention enhances the auditory

40-Hz transient response in humans Nature 364, 59–60

h Joliot, M. et al. (1994) Human oscillatory brain activity near 40 Hz

coexists with cognitive temporal binding Proc. Natl. Acad. Sci.

U. S. A. 91, 11748–11751

i Tallon, C. et al. (1995) Gamma-range activity evoked by coherent

visual stimuli in humans Eur. J. Neurosci. 7, 1285–1291

j Sannita, W.G. et al. (1995) Scalp-recorded oscillatory potentials

evoked by transient pattern-reversal visual stimulation in man

Electroencephalogr. Clin. Neurophysiol. 96, 206–218

k Jokeit, H. et al. (1994) in Oscillatory Event-Related Brain Dynamics

(Pantev, C., Elbert, T. and Lutkenhoner, B., eds), pp. 135–146,

Plenum Press

l Tallon-Baudry, C. et al. (1997) Combined EEG and MEG recordings

of visual 40 Hz responses to illusory triangles in human

NeuroReport 8, 1103–1107

m Lopez, L. and Sannita, W.G. (1997) Magnetically recorded

oscillatory responses to luminance stimulation in man


n Hashimoto, I. et al. (1996) Somatic evoked high-frequency

magnetic oscillations reflect activity of inhibitory interneurons in

the human somatosensory cortex Electroencephalogr. Clin.

Neurophysiol. 100, 189–203

o Curio, G. et al. (1997) Somatotopic source arrangement of 600 Hz

oscillatory magnetic fields at the human primary somatosensory

hand cortex Neurosci. Lett. 234, 131–134

p Leonards, U. et al. (1996) The influence of temporal phase

differences on texture segmentation Vis. Res. 36, 2689–2697

q Alais, D. et al. (1998) Visual features that vary together over time

group together over space Nat. Neurosci. 1, 160–164

r Usher, M. and Donnelly, N. (1998) Visual synchrony affects binding

and segmentation in perception Nature 394, 179–182

s Fahle, M. and Koch, C. (1995) Spatial displacement, but not

temporal asynchrony, destroys figural binding Vis. Res. 35,

491–494

t Kiper, D.C. et al. (1996) Cortical oscillatory responses do not affect

visual segmentation Vis. Res. 36, 539–544

u Regan, D. and Spekreijse, H. (1986) Evoked potentials in vision

research 1961–86 Vis. Res. 26, 1461–1480

v Galambos, R. et al. (1981) A 40-Hz auditory potential recorded

from the human scalp Proc. Natl. Acad. Sci. U. S. A. 78, 2643–2647



Evoked gamma activity is strictly phase-locked to the stimulus onset and can beseen in the averaged evoked response, whereas the latency of induced gamma ac-tivity varies across trials and is therefore canceled out by averaging. Specific meth-ods are thus necessary to detect and characterize these induced gamma responses.

One simple possibility is to estimate, on each single-trial, the time-varia-tion of gamma power in a particular frequency band around 40 Hz (Ref. a).However, as latency and frequency of gamma activities are not known a pri-ori, a time-frequency approach should be preferred. This can be achieved bycomputing the time-varying spectra of the EEG tapered by a moving windowof fixed duration (Ref. b) (Gabor’s transform). We propose an alternativemethod that estimates the time–frequency power of the signals by means of acomplex Morlet’s wavelet transform (Fig. F,G) which provides a better com-promise between time and frequency resolutions (Ref. c).

When this analysis is applied to the averaged response, the phase-lockedgamma activity can be clearly identified (Fig. C). When this method is ap-plied to each single trial, followed by an averaging of the powers across trials(Fig. E), it becomes possible to identify non-phase-locked activity as long asthe signal-to-noise ratio is high enough and the jitter does not exceed thewavelet duration.

References

a Pfurtscheller, G. et al. (1993) 40-Hz oscillations during motor behavior in man

Neurosci. Lett. 164, 179–182

b Makeig, S. (1993) Auditory event-related dynamics of the EEG spectrum and

effects of exposure to tones Electroencephalogr. Clin. Neurophysiol. 86, 283B293

c Sinkkonen, J. et al. (1995) Gabor filters: an informative way for analysing event-

related brain activity J. Neurosci. Methods 56, 99–104

Box 2. How to identify evoked and induced gamma activities

Fig. I. (A) Successive EEG trials(simulated data) with a smallamplitude gamma responsephase-locked to stimulus onset(blue boxes) and a gamma burstjittering in latency (greenboxes). (B) Averaging acrosssingle trials leads to the conven-tional evoked potential. (C)Time–frequency power repre-sentation of the evoked gammaresponse. The abscissa repre-sents time, and the ordinate,frequency. The color scale codesthe variations of power (posi-tive or negative) with respect toa pre-stimulus baseline. Thenon-phase-locked activity iscancelled out. When the time-frequency power is computedfor each singe trial (D), andthen averaged across trials (E),the induced gamma response isrevealed. (F, G) Principle of thewavelet transform. (F) The sig-nal to be analyzed s(t) is con-voluted by complex Morlet’swavelets w(t,f) having aGaussian shape both in the timedomain (standard deviation σt),and in the frequency domain(σf=1/(2πσt)) around a frequencyf : w ( t , f ) = A . e x p ( – t 2 / 2 σ t

2 ) .exp(2iπft) with A being a nor-malization factor: A=1/(σt √π)1/2.As opposed to the classical moving Fourier transform, awavelet family is defined by aconstant ratio f/σf (greater thanfive in practice). The waveletstherefore have the same num-ber of cycles for different fre-quency bands, resulting in dif-ferent wavelet durations. Thetime-varying power of the signalaround frequency f is thesquared modulus of the con-volution: P(t,f) = |w(t,f)*s(t)|2.(G) Repeating this calculationfor a family of wavelets, havingdifferent frequencies f, pro-vides a time–frequency powerrepresentation of the signalcomponents.

areas. An increase in gamma activity has also been reportedprior to, or during, tasks requiring a high level of sensori-motor integration both in human26 and monkeys27. Thistype of gamma activity has therefore been proposed to be related to the integration of sensory and motor processesduring movement preparation and/or control of movement.

Detection tasksInduced gamma activity in the 40 Hz range appeared in thescalp EEG, 200 ms after electrical stimulation of the fingerduring a sensory discrimination task28. However, in a simi-lar task, others did not find any inter-electrode correlationsin the gamma-range from intracranial recordings over thesomatosensory areas29. This latter negative result might bebecause this study looked for coherence over a relativelylong time window (500 ms) compared with the rather transient (around 125 ms) inter-electrode phase-locking re-sponse in the scalp EEG. During a simple auditory detec-tion task30, the strength of the induced 40-Hz activity in-creased between 200 and 400 ms after stimulus onset insubjects who reacted rapidly, whereas it increased beforestimulus onset for subjects who reacted more slowly.

Modulation by attentionIn early experiments, intracranially recorded occipital gammaresponses to static visual stimuli decreased in amplitude whenthe subject did not pay attention to a picture that was pre-sented31. Various sensory stimuli (light-flash, speech, cold,

etc.) induced a gamma oscillatory response 100Ð200 ms afterstimulus onset in the posterior temporo-parietal lobe with atendency for rapid habituation32. More recently, in passive andactive auditory paradigms differences have been reported inthe EEG gamma strength 200 ms and 300 ms after subjectsheard standard tones compared to when they heard devianttones33. In similar situations34, we have observed that the time-course of the induced 40-Hz activity is characterized by a re-duction (at around 100 ms) followed by an increase (peakingat around 250 ms) compared with pre-stimulus baseline ac-tivity (Fig. 1). This temporal pattern is similar to that foundover the auditory cortex in the anaesthetized rat35. During attentive listening, the increase in gamma strength was en-hanced and prolonged34, and its scalp topography was differ-ent from that of the auditory evoked components known toinvolve an activity in both auditory cortices (Fig. 1C and 1D).

Complex tasksIn contrast to the studies mentioned above, which reportedgamma activity that was induced by a sensory stimulus, sev-eral authors have reported global increases in gamma signalstrength during complex tasks compared to control situa-tions. For example, Spydell and Sheer36 found an increasedgamma signal that was left-lateralized during a verbal task,and was right-lateralized during a visual rotation task. Theysuggested that the gamma activity could reflect a ‘focussedarousal’ in the task-relevant neural circuitry. A gamma en-hancement has also been observed at frontal sites37 during



Opinion

Fig. 1. Acoustically induced gamma activity and selective attention. (A) Passive listening. Acoustic stimuli (black bar: 1000 Hz, 50ms duration tone-bursts) were delivered binaurally to subjects who were watching a silent video movie. The average time-frequencypower distribution at vertex electrode (grand average over 10 subjects) shows variations in induced gamma activity with respect to thepre-stimulus baseline: an early decrease (red) around 40 Hz and 80 ms after stimulus onset, followed by increases (yellow) around 45 Hzpeaking at 210 and 400 ms. (B) Active listening. Average time-frequency power distribution at vertex electrode after the same acousticstimuli (black bar: 1000 Hz, 50 ms duration tone-bursts) while the subjects are detecting 1050-Hz rare tones. This distribution shows alarger gamma power increase (yellow) compared to the passive listening situation, peaking later (300 ms) and having a longer duration.Auditory induced gamma activity is thus enhanced by selective attention. (C) The scalp topography of the induced gamma response at 210 ms has a maximum at parietal electrodes, whereas (D) the major auditory evoked components between 30 and 180 ms, known to involve both auditory cortices, show a different topography, namely, a polarity reversal pattern delineating the Sylvian fissure. (See Ref. 34.)

A Passive listening B Active listening

C Induced gamma D Auditory evoked components

multi-stable visual perception and was suggested to be re-lated to phases of frequent perceptual switching, rather thanphases of stable perception.

Common properties of induced gamma activityFrom this brief overview it appears that stimulus-inducedvariations in gamma activity occur within similar time-ranges (200Ð400 ms) in various sensory modalities andtasks. The scalp topography of this induced gamma activitywas not systematically investigated, nevertheless, it does notseem to reflect a global and unspecific activation of ex-tended cortical structures. For example, in the motor tasksthe somatotopic mapping of the movement-related gammaresponse suggests a rather restricted area of activation. Inmany of these studies, the induced gamma activity was in-terpreted, in very general terms, as the neural substrate of‘cognitive processes’. This interpretation is difficult to relate

to any precise functional hypothesis, al-though a common finding seems to be thatinduced gamma activity increases duringcomplex or attention-demanding tasks.The next sections of this article will ex-amine in more detail whether inducedgamma activity could underlie the re-presentational hypothesis. We will firstpresent experimental results specificallyrelated to bottom-up feature-bindingprocesses, and then, those supporting anotion of active representation involvingtop-down processes.

Induced gamma and bottom-up feature-bindingIf induced gamma activity reflects a bind-ing mechanism, it should be enhancedwhen a coherent percept is created in re-sponse to a given stimulus. This hypoth-esis can be tested by evaluating the strengthof the gamma signal that is induced bystimuli that share the same physical prop-erties but do, or do not, lead to the per-ception of a coherent percept. Using pro-tocols taken from animal studies9, twodifferent groups have reported an increasein the strength of the gamma signal in thescalp EEG while subjects were passivelyviewing coherent moving patterns38Ð40.The topography of this 40-Hz increasematched the retinotopic organization ofareas V1 and V2 in humans38. These re-sults also provide evidence for the func-tional specificity of gamma activity withrespect to alpha activity39,40.

Increases of gamma activity have alsobeen reported in response to coherent versus incoherent static stimuli in activetasks. We analyzed the EEG in responseto a coherent and a non-coherent visualstimulus in a discrimination task41. Ashort-lasting occipital enhancement in

the 30Ð50 Hz band was observed around 280 ms after stim-ulus onset in response to coherent triangles only (Fig. 2).No such effect was observed at higher frequencies (up to100 Hz), or in the averaged evoked potential. Only the in-duced gamma activity varied consistently in line with ourhypothesis. Similarly, a peak of induced gamma activity wasreported at 35Ð40 Hz, 230 ms after the presentation ofhuman (Mooney) faces42. This response was much larger inresponse to pictures presented in the upright position thanfor inverted pictures (with no perceived facial content). Thestereoscopic fusion of random-dot stereograms into a full 3-D percept, compared with a 2-D perception con-dition, also elicited an occipital gamma increase in theEEG43, while no difference was observed in the alpha orbeta bands.

These results provide evidence for a role for inducedgamma activity in the construction of coherent representations



Fig. 2. Induced gamma activity and visual bottom-up feature binding. (A) Subjects were presented withfour different stimuli. Both the illusory (Kanizsa) and the real triangles were coherent stimuli, leading to a coherent percept through a bottom-up feature binding process. The ‘no-triangle’ stimulus served as control.Subjects had to silently count the occurrences of the target stimulus, a curved illusory triangle, and to report thisnumber at the end of each recording block. This task, when correctly performed, ensured that subjects perceivedcorrectly illusory triangles and remained attentive throughout the whole recording session. (B) Time–frequencypower of the EEG at electrode Cz (overall average of 8 subjects), in response to the illusory triangle (top) and tothe no-triangle stimulus (bottom), the baseline level being taken in the pre-stimulus period. Two successive burstsof oscillatory activities were observed. A first burst occurred at about 100 ms and 40 Hz; it was an evoked re-sponse, phase-locked to stimulus onset, peaking at electrode Cz. It showed no difference between stimulus types,and thus cannot reflect the spatial feature-binding process required to perceive the triangles. The second burst,around 280 ms and between 30 and 60 Hz, was induced by the stimulus, and most prominent in response to co-herent stimuli. There was no statistical difference in the gamma range between the responses to the illusory andreal triangles. Induced gamma could thus reflect the spatial binding of the elementary features of the picture intoa coherent representation of the triangle. It should be noted that no component of the evoked response dis-criminated between coherent and non-coherent stimuli. (C) Topography of the gamma power averaged between250 and 350 ms and 30 and 60 Hz, in response to the illusory triangle. Maximum activity is observed at occipitalelectrodes. (See Ref. 41.)

based on the integration of visual infor-mation. The signal is enhanced when spa-tially distinct features have to be boundtogether, compared with when they remainsegregated as distinct percepts. No modi-fication of evoked potentials or alpharhythm was found to be related to thishypothesized feature-binding process.Moreover, in response to a coherent vi-sual stimulus, the induced gamma activitywas maximal at occipital and parieto-occipital locations, which suggests that itoriginates, at least in part, in visual areas.This view is also supported by the findingthat it partially follows a retinotopicorganization (upper/ lower hemifields).

The role of 40 Hz activity has alsobeen explored in language studies inwhich it has been hypothesized that theactivity could reflect associations betweenwords and meanings. Indeed, a differencehas been observed at long latencies (morethan 400 ms) in the strength of thegamma signal between words andpseudo-words both in the visual (EEG)44

and auditory (MEG45 and EEG46) modal-ities. Furthermore, the topography ofgamma activity induced by visually pre-sented nouns and verbs revealed differ-ences between 500 and 800 ms47. Theseresults suggest that induced gamma couldbe related to the visual or motor associ-ations prompted by verbal stimuli, al-though the long latencies of the effectsobserved seem to indicate that they donot reflect a primary differential process-ing of words and pseudo-words.

Induced gamma and objectrepresentationAs outlined in the introduction, an objectrepresentation can be directly built fromthe sensory input by bottom-up processes,but might also be expected to be activated,retrieved, or rehearsed through top-downprocesses. To examine the latter, we inves-tigated the variations in gamma strengththat occur when an internal represen-tation of a picture or a tone is needed tocorrectly perform a task. When searchingfor someone in a crowd, we know in ad-vance who we are looking for – in otherwords, we generate an internal representa-tion of the particular person we are search-ing for. A similar process is involved whena subject must detect a hidden object in a picture. In our ex-periment48, we used a modified version of the well-knownDalmatian dog picture (Fig. 3). Naive subjects did not per-ceive the hidden dog, and measurement revealed only a weakinduced gamma response around 280 ms after stimulus

onset. Once the subjects were trained to perceive and detectthe hidden dog, a much larger induced gamma response wasrecorded, irrespective of whether the stimulus contained aDalmatian dog or not. Our interpretation of this result isthat the activity may reflect the top-down activation of a



Opinion

Fig. 3. Induced gamma activity and top-down activation of an internal representation. The experimentwas divided in two conditions. In the first one (A, top row), naive subjects were presented with three differentstimuli. Two of them (neutral stimulus and unperceived dog) were perceived as meaningless black blobs on a graybackground. The task of the subject was to count the occurrences of the ‘twirl’ target stimulus. In this condition,only a very faint induced gamma response could be observed (B, left). Nevertheless, the peaking power of thisgamma response was enhanced after the target stimulus (C). This may reflect the bottom-up feature bindingprocess involved in the twirl perception. Subjects were then trained to perceive the hidden Dalmatian dog, withits head oriented to the left or to the right (A, bottom row); (the outlines of the dog are presented below thestimuli to help the reader to detect them in the pictures). In this second condition, their task was to count the oc-currences of the Dalmatian dog with its head oriented leftward. Compared to the first condition, a large increasein the induced gamma response was observed in response to all stimuli (B and C). This increase could reflect thetop-down activation of the internal representation of the attended Dalmatian dog. Indeed, the subject neededthis representation to perform the task, independently of the stimulus presented. An additional increase of theinduced gamma response was observed in response to the target dog, and in a lesser extent to the dog with itshead oriented to the right. This may reflect the bottom-up process of linking the black blobs into a coherent dogpicture. The topography of the induced response in the second condition is shown in (D). It peaks at occipital andright parietal locations. Differences between stimuli and/or conditions could also be found in the evoked poten-tials, but they all occurred at longer latencies, later than the variations of the induced gamma response. (See Ref. 48.)

representation of the object (the dog) during the visualsearch. An increase in activity was also observed when theblack blobs in the picture were linked together to build a co-herent picture (bottom-up process). This gamma effect oc-curred 10 ms earlier than the first difference observed in theaveraged evoked potentials.

Internal representations may also bedriven during rehearsal processes withinshort-term memory. In the visual modal-ity, we studied the gamma responseselicited by a delayed-matching-to-sampletask49 (Fig. 4A). Sustained oscillatory ac-tivities in the gamma bands, and also inthe beta (15Ð20 Hz) bands were observedduring the delay period, when the targetstimulus was presumably being rehearsed.These activities disappeared in a controlcondition, where no memory was re-quired (Fig. 4B). We therefore suggestedthat the activity during the delay periodcould reflect the maintenance of a visualobject representation in short-term mem-ory. The topography of this activity, indi-cated a contribution from occipito-tem-poral as well as frontal areas, and isconsistent with the known functionalanatomy of short-term memory networksin humans50. No other component of theevoked potential lasted long enough to belikely to reflect the rehearsal of the stimu-lus representation in memory51.

A possible role for induced gammaactivity in acoustic object representationwas tested in a difficult frequency dis-crimination paradigm34. Standard toneswere presented, randomly intermixedwith two types of deviant tones of slightlyhigher frequency, one of which was thetarget stimulus that was to be detected.An increase in induced gamma activityaround 40 Hz was found, starting 250 msafter the onset of standard tones (Fig. 5).This gamma activity was clearly pro-longed (up to 800 ms) when the subjectswere expecting a target to occur againsoon. Such sustained gamma activity afterstandard tones could be interpreted interms of rehearsal of the pitch represen-tation required to perform the discrimi-nation task correctly. This might be relatedto the strategy reported by the subjectsthat they were internally ‘re-playing’ thestandard tone in order to be able to detectthe target. However, an interpretation interms of increased attention can not bediscounted in this experiment.

Induced gamma oscillations: thesignature of object representation?In all the studies described above, the

variations of induced gamma activity are predicted by therepresentational hypothesis. Thus, there is growing experi-mental evidence pointing toward a functional role for in-duced gamma activity in binding together the areas involvedin an object representation, whether generated through bottom-up or top-down processes. Alternative explanations



Fig. 4. Induced gamma activity and the rehearsal of a visual object representation in short-term mem-ory. (A) In the memory condition, subjects performed a delayed-matching-to-sample task: a first stimulus(smooth shape) was presented, followed by a delay and a second stimulus. Subjects had to press a button whenthe two stimuli were exactly identical. To perform the task, the representation of the first stimulus had to bemaintained in short-term memory. The time–frequency power at electrode O1 showed three successive bursts ofactivity in the gamma range: at 280 ms (ON response), 700 ms (OFF response), and later during the delay. Thetopography of the gamma power during the delay showed an additional bilateral frontal component as com-pared to the occipital topography of the ON response. (B) In the control condition, no memory was required: sub-jects had to detect a possible dimming of the fixation point at the end of the delay. The ON induced gamma response at 280 ms was reduced, and the activity during the delay completely suppressed. The gamma activityduring the delay could thus reflect the rehearsal of the first stimulus in short-term memory. Furthermore, itshowed specific variations and/or topography as compared to the alpha and beta band, and to the evoked potentials. (See Ref. 49.)

can be found for each experiment individually (attentionalselection of the stimuli looking like the target in the triangle experiment, for example), but none of the alternative accounts can explain the whole set of results.

Furthermore, induced gamma activity was the onlycomponent to vary systematically according to the represen-tational hypothesis. Its functional specificity was demon-strated in several studies reported here, mainly in the visualmodality. Induced gamma activity was topographically andtemporally distinct from the alpha rhythm and from theevoked low frequency or gamma components, and thuscannot be considered as an epiphenomenon of any of thesecomponents. In particular, to answer the question raised byJürgens et al.52, it does not seem to reflect a harmonic ofalpha activity40,49. However, its functional relationship toactivities in the beta (15–20 Hz) range remains un-clear43,49,51. Finally, the variations in induced gamma top-ography should be emphasized. Different areas seem to beinvolved in different tasks. This could indicate the existenceof a flexible mechanism of dynamic recruitment of func-tional areas in oscillatory ensembles.

Neural substrate of induced gamma activityWhere do these oscillations recorded on the scalp originate?One could argue that muscular activity might account forthese signals. However, several arguments can be put for-ward that rule out this interpretation. Since induced gammaactivity shows task-dependent time courses and topogra-phies, it seems unlikely that it could reflect muscle activityalone. Furthermore, the functional effects are confined to afairly narrow frequency range (30Ð50 Hz), whereas muscle

activity usually has a broader spectral content49. Moreover,visually evoked and induced gamma bursts have been ob-served intracranially in human at the same latencies as thoseobserved on the scalp surface53.

In order to identify the underlying neural source of theinduced gamma oscillations, we repeated the Kanizsa tri-angle experiment (Fig. 2) and recorded simultaneouslyEEG and MEG signals over occipital areas54. The transient40 Hz evoked response at 100 ms was detected both in EEGand MEG recordings, whereas the induced 30Ð60 Hzgamma activity at 250 ms was observed in the EEG only.Preliminary results in the auditory modality showed a simi-lar difference between EEG and MEG in induced gammaactivity during passive listening. This absence of gamma activity in the MEG signal could be explained by deep orradially oriented neural sources which gives rise to a weakmagnetic field, or by an unusual source configuration.

In support of this latter interpretation, we proposed aring-shaped distribution of current dipoles (Fig. 6) as asource model of the induced gamma oscillations55, whichcreates an electric potential field but not a magnetic field (oronly a weak one) on the scalp surface. Moreover, it gener-ates a field potential that does not reverse through corticaldepth, as has been observed in animal studies with sponta-neous56 or stimulus-induced gamma activities11. A ring-shaped source model could mimic the activity of synapseslocated on horizontally oriented dendrites. This geometry isconsistent with the idea that interneurons might be involvedin a network generating a coherent oscillatory activity57Ð59.However, a possible contribution of pyramidal neurons(basal dendrites for example) cannot be ruled out60.



Opinion

Fig. 5. Induced gamma activity and sound representation. (A) Acoustic stimuli (pure tone-bursts, 50 ms duration) were deliveredbinaurally with a constant inter-stimulus interval (1.4 s): frequent standard tones at 1000 Hz (black bars) and rare deviant tones, targetat 1040 Hz (red bars) or distractor at 1080 Hz (brown bars). Subjects had to press a button after each target tone, and they were awarethat at least 3 standards were delivered between two deviants. The tone frequencies were chosen in order to make the task difficultenough so that it could not be performed automatically (85% of correct responses). (B) The average time–frequency power distributionswere analyzed separately for the second standard following a deviant (second STD, in blue) and for the standard preceding a deviant (lastSTD, in green). (C) Different time courses of the gamma power were induced by the same acoustic stimulus: the gamma response peakedsignificantly later for the last STD (440 ms) than for the second STD (210 ms), and persisted for longer (up to 800 ms). Furthermore, fol-lowing deviant tones, i.e. target (in red) or distractor (in brown), the gamma activity did not show any post-stimulus increase but rathera transient decrease followed by a return to baseline. The prolonged induced gamma activity after several standard tones could be in-terpreted in terms of rehearsal of the tone pitch representation required to correctly perform the task, while the absence of gamma increase after deviant tones suggests that the frequency discrimination itself is performed by faster processes and no sound rehearsal isfurther required. This representation is more active when the subject anticipates that a deviant is about to occur. Nevertheless, an inter-pretation in terms of increasing attention cannot be discounted. (See Ref. 34.)

Gamma oscillations and spike synchronizationIn order to better understand the neural substrate of inducedgamma oscillations, we still need to fill the gap between themacroscopic approaches based on scalp EEG recordings re-viewed here, and the spike-train synchronizations observedin animal studies at a microscopic scale in single or multiple-unit recordings14,16,61,62. At the present time the results ofthese two approaches are difficult to compare because of thevery different spatial sampling and experimental paradigmsused. Signal analysis methods are also very different. In ani-mal studies, spike synchronizations are evaluated over quitelong time periods while, in human EEG or MEG, variationsin gamma strength are usually estimated over much shorterperiods. The rather short-lived variations in inducedgamma strength observed in humans could easily be ob-scured in cross-correlograms computed over several hun-dreds of milliseconds. Indeed, some authors have found noevidence for oscillatory synchronization in monkey cortex63Ð65

and this may be partly due to the use of exceedingly longtime windows for computing auto- or cross-correlograms66,systematic sampling bias toward a particular layer67 or dif-ferences in recording techniques. In addition, part of thecontroversy in animal studies is due to the distinction be-tween oscillatory patterns in the spike train of a single neuron and oscillatory synchronization between neur-ons66,68. It is worth noting that whenever an oscillatory ac-tivity emerges in the scalp EEG, a certain level of underly-ing neural synchronization is necessarily reached.

Nevertheless, the gamma activities observed at differentrecording levels share a number of properties since they arenot phase-locked to stimulus onset, and they are enhancedin response to coherent stimuli. An analysis at intermediatelevels (namely local-field potentials or electro-corticograms),using similar experimental paradigms could be a way tobridge this gap. Indeed, the time course of visually oracoustically induced gamma activity observed in humanEEG is comparable to that found in cat local-field poten-tials69 or rat electro-corticograms35. Finally, synchronizationbetween EEG leads could also be evaluated on the scalp surface or in human intracranial recordings53. However, thecontribution of a single neural source located at a distancefrom the two testing electrodes should first be ruled out be-cause it could create an artificial synchronization of EEGsignals by simple volume conduction effect.

ConclusionThe representational hypothesis presented here is derivedfrom the feature-binding hypothesis from the animal litera-ture and postulates that fast oscillatory synchronization ofbrain areas underlies the construction of a task-relevant object representation. When searching for experimental evidence in human EEG/MEG recordings to support thishypothesis, two types of oscillatory activity in the gamma-range should be distinguished (1) an early, transient 40-Hzevoked response and (2) a stimulus-induced response atlonger latencies. Only the induced gamma activity, as op-posed to the evoked activity (see Box 1), seems to be relatedto the generation of an object representation, both in visualand auditory modalities. In addition, the functional varia-tions in this signal, at least in the visual modality, are differ-ent from those of the alpha rhythm and of averaged evokedlow frequency or gamma components.

Induced gamma activities show a high degree of spatialand temporal flexibility that depends upon the sensory sys-tems recruited and the experimental tasks involved. Theyshould not be considered as a unique or stereotyped brainresponse derived from the same set of neural sources. Onthe contrary, they seem to reflect interactions within net-works organized both in space and time. This accords withan interpretation in terms of dynamic cooperative mecha-nisms that link together various brain areas that encode thedifferent features of an object and have their own specificfunction in relation to the task to be performed. Fast oscil-latory synchronization of distributed cell assemblies maythus be proposed as a very general neural mechanism un-derlying sensory integration and object representation.However, the transition from the emergence of such assemblies to behavior is not clearly understood at present.



Fig. 6. Possible source model of induced gamma activity. (A) An equivalent currentdipole source model, oriented perpendicular to the cortical surface, is usually proposed forthe pyramidal cell synaptic activity underlying the early evoked components. This sourcemodel fits with the polarity reversal often observed in cortical depth for such components.It generates a recordable scalp potential distribution, and a magnetic field if oriented par-allel to the scalp surface (along folded sulci for example). (B) A ring-shaped distribution ofdipoles is proposed as a possible source model for the induced gamma response. This con-figuration creates a field potential that does not reverse through cortical depth as observedin some animal studies. Moreover, it generates a potential distribution, but not a magneticfield on the scalp surface (or a weak one only), which corresponds to our observation thatstimulus-induced gamma oscillations are detectable in EEG data, but not in simultaneouslyrecorded MEG data. Such ring-shaped sources could mimic the activity of synapses locatedon horizontally oriented dendrites distributed all around the soma. This geometry fits withthe idea that interneurons might be involved in a network generating a coherent oscillatoryactivity. (See Ref. 55.)

Outstanding questions

• What is the functional significance of the transient decrease of gammastrength observed after a sensory stimulus, in the time range where themajor components of the evoked responses occur?

• Is multi-sensory integration also achieved through oscillatorysynchronization ?

• Induced gamma activity seems to underlie the activation of fine objectrepresentations. Can more crude representations be established morequickly through other mechanisms ?

• Is it possible to bridge the gap between oscillatory synchronizationobserved at microscopic levels in animals and modulation of gammastrength observed at a macroscopic scale in humans?

• Could the sometimes long-lasting induced gamma responses bettercorrelate with metabolic or haemodynamic images (PET or fMRI) thanthe brief components of the evoked potential ?

While early evoked potentials could reflect the encoding ofsome physical attributes of the stimulus, the inducedgamma oscillations that follow might play a crucial role inbinding together those areas that are necessary to perform aparticular task. Different types of neurons and networksmight underlie each of these electrophysiological compo-nents. How these two mechanisms combine, and which onedominates in a given task, remains an open question.

Acknowledgements

This work was supported by grants from Human Frontier Science

Program (1995–1998) and the Rhône-Alpes Region (1997–2000). We

thank J. Pernier for his contribution to the neural source model and

many helpful discussions, C. Pantev from Münster University for his

collaboration in MEG recordings, M. Huotilainen for her comments on

the manuscript, and J.F. Echallier and P.E. Aguerra for their helpful

technical assistance.

References

1 Courtney, S.M. and Ungerleider, L.G. (1997) What fMRI has taught us

about human vision Curr. Opin. Neurobiol. 7, 554–561

2 Posner, M.I. and Raichle, M.E. (1994) Images of Mind, Scientific

American Library

3 Van Essen, D.C. et al. (1992) Information processing in the primate

visual system: an integrated systems perspective Science 255, 419–423

4 Orban, G.A. et al. (1996) Task dependency of visual processing in the

human visual system Behav. Brain Res. 76, 215–223

5 Delage, Y. (1919) Le reve: Etude Psychologique, Philosophique et

Littéraire, Presses Universitaires de France

6 Milner, P.M. (1974) A model for visual shape recognition Psychol. Rev.

81, 521–535

7 von der Malsburg, C. and Schneider, W. (1986) A neural cocktail-party

processor Biol. Cybern. 54, 29–40

8 Damasio, A.R. (1989) The brain binds entities and events by

multiregional activation from convergence zones Neural Comput. 1,

123–132

9 Singer, W. and Gray, C.M. (1995) Visual feature integration and the

temporal correlation hypothesis Annu. Rev. Neurosci. 18, 555–586

10 Brosch, M. et al. (1997) Stimulus-dependent modulations of correlated

high-frequency oscillations in cat visual cortex Cereb. Cortex 7, 70–76

11 Eckhorn, R. et al. (1988) Coherent oscillations: a mechanism of feature

linking in the visual cortex?: multiple-electrode and correlation

analyses in the cat Biol. Cybern. 60, 121–130

12 Engel, A.K. et al. (1991) Synchronization of oscillatory neuronal

responses between striate and extrastriate visual cortical areas of the

cat Proc. Natl. Acad. Sci. U. S. A. 88, 6048–6052

13 Freiwald, W.A. et al. (1995) Stimulus dependent intercolumnar

synchronization of single unit responses in cat area 17 NeuroReport 6,

2348–2352

14 Gray, C.M. et al. (1989) Oscillatory responses in cat visual cortex exhibit

inter-columnar synchronization which reflects global stimulus

properties Nature 338, 334–337

15 Frien, A. et al. (1994) Stimulus-specific fast oscillations at zero phase

between visual areas V1 and V2 of awake monkey NeuroReport 5,

2273–2277

16 Kreiter, A.K. and Singer, W. (1996) Stimulus-dependent

synchronization of neuronal responses in the visual cortex of the

awake macaque monkey J. Neurosci. 16, 2381–2396

17 Barinaga, M. (1990) The mind revealed? Science 249, 856–858

18 Singer, W. et al. (1997) Neuronal assemblies : necessity, signature and

detectability Trends Cognit. Sci. 1, 252–261

19 Pulvermuller, F. et al. (1997) High-frequency brain activity: its possible

role in attention, perception and language processing Prog.

Neurobiol. 52, 427–445

20 Pfurtscheller, G. et al. (1994) Differentiation between finger, toe and

tongue movement in man based on 40 Hz EEG Electroencephalogr.

Clin. Neurophysiol. 90, 456–460

21 Salenius, S. et al. (1996) Human cortical 40 Hz rhythm is closely related

to EMG rhythmicity Neurosci. Lett. 213, 75–78



Opinion

22 Brown, P. et al. (1998) Cortical correlate of the piper rhythm in humans

J. Neurophysiol. 80, 2911–2917

23 Bouyer, J.J. et al. (1987) Anatomical localization of cortical beta

rhythms in cat Neuroscience 22, 863–869

24 Sanes, J.N. and Donoghue, J.P. (1993) Oscillations in local field

potentials of the primate motor cortex during voluntary movement

Proc. Natl. Acad. Sci. U. S. A. 90, 4470–4474

25 Donoghue, J.P. et al. (1998) Neural discharge and local field potential

oscillations in primate motor cortex during voluntary movements

J. Neurophysiol. 79, 159–173

26 Kristeva-Feige, R. et al. (1993) Oscillatory brain activity during a motor

task NeuroReport 4, 1291–1294

27 Murthy, V.N. and Fetz, E.E. (1992) Coherent 25- to 35-Hz oscillations in

the sensorimotor cortex of awake behaving monkeys Proc. Natl. Acad.

Sci. U. S. A. 89, 5670–5674

28 Desmedt, J.E. and Tomberg, C. (1994) Transient phase-locking of 40 Hz

electrical oscillations in prefrontal and parietal human cortex reflects

the process of conscious somatic perception Neurosci. Lett. 168, 126–129

29 Menon, V. et al. (1996) Spatio-temporal correlations in human gamma

band electrocorticograms Electroencephalogr. Clin. Neurophysiol. 98,

89–102

30 Jokeit, H. and Makeig, S. (1994) Different event-related patterns of

gamma-band power in brain waves of fast- and slow-reacting subjects


31 Sem-Jacobsen, C.W. et al. (1956) Electroencephalographic rhythms

from the depths of the parietal, occipital, and temporal lobes in man


32 Perez-Borja, C. et al. (1961) Depth electrographic studies of a focal fast

response to sensory stimulation in the human Electroencephalogr.

Clin. Neurophysiol. 13, 695–702

33 Marshall, L. et al. (1996) Event-related gamma band activity during

passive and active oddball tasks NeuroReport 7, 1517–1520

34 Bertrand, O. et al. (1998) Auditory induced 40-Hz activity during a

frequency discrimination task NeuroImage 7, S370

35 Franowicz, M.N. and Barth, D.S. (1995) Comparison of evoked

potentials and high-frequency (gamma-band) oscillating potentials in

rat auditory cortex J. Neurophysiol. 74, 96–112

36 Spydell, J.D. and Sheer, D.E. (1982) Effect of problem solving on right

and left hemisphere 40 hertz EEG activity Psychophysiology 19,

420–425

37 Basar-Eroglu, C. et al. (1996) Frontal gamma-band enhancement

during multistable visual perception Int. J. Psychophysiol. 24, 113–125

38 Lutzenberger, W. et al. (1995) Visual stimulation alters local 40-Hz

responses in humans: an EEG study Neurosci. Lett. 183, 39–42

39 Muller, M.M. et al. (1996) Visually induced gamma-band responses in

human electroencephalographic activity: a link to animal studies Exp.

Brain Res. 112, 96–102

40 Muller, M.M. et al. (1997) Visually induced gamma-band responses to

coherent and incoherent motion: a replication study NeuroReport 8,

2575–2579

41 Tallon-Baudry, C. et al. (1996) Stimulus specificity of phase-locked and

non-phase-locked 40 Hz visual responses in human J. Neurosci. 16,

4240–4249

42 Rodriguez, E. et al. (1998) Meaningful visual perception induces

complex spatio-temporal patterns of phase synchrony on scalp

recordings NeuroImage 7, S344

43 Revonsuo, A. et al. (1997) The neural generation of a unified illusion in

human vision NeuroReport 8, 3867–3870

44 Lutzenberger, W. et al. (1994) Words and pseudowords elicit distinct

patterns of 30-Hz EEG responses in humans Neurosci. Lett. 176,

115–118

45 Pulvermuller, F. et al. (1996) High-frequency cortical responses reflect

lexical processing: an MEG study Electroencephalogr. Clin.


46 Krause, C.M. et al. (1998) Automatic auditory word perception as

measured by 40 Hz EEG responses Electroencephalogr. Clin.


47 Pulvermuller, F. et al. (1996) Brain rhythms of language: nouns versus

verbs Eur. J. Neurosci. 8, 937–941

48 Tallon-Baudry, C. et al. (1997) Oscillatory gamma-band (30–70 Hz)

activity induced by a visual search task in humans J. Neurosci. 17,

722–734

49 Tallon-Baudry, C. et al. (1998) Induced gamma-band activity during the

delay of a visual short-term memory task in humans J. Neurosci. 18,

4244–4254

50 Ungerleider, L.G. (1995) Functional brain imaging studies of cortical

mechanisms for memory Science 270, 769–775

51 Tallon-Baudry, C. et al. Sustained and transient oscillatory responses in

the gamma and beta bands in a visual short-term memory task in

humans Vis. Neurosci. (in press)

52 Jürgens, E. et al. (1995) Stimulus-induced gamma oscillations:

harmonics of alpha activity? NeuroReport 6, 813–816

53 Lachaux, J.P. et al. (1998) Synchrony in gamma-band oscillations in

human intracortical recordings during visual discrimination

NeuroImage 7, S303

54 Tallon-Baudry, C. et al. (1997) Combined EEG and MEG recordings of

visual 40 Hz responses to illusory triangles in human NeuroReport 8,

1103–1107

55 Tallon-Baudry, C. et al. A ring-shaped distribution of dipoles as a

source model of induced gamma-band activity Electroencephalogr.

Clin. Neurophysiol. (in press)

56 Steriade, M. and Amzica, F. (1996) Intracortical and corticothalamic

coherency of fast spontaneous oscillations Proc. Natl. Acad. Sci. U. S. A.

93, 2533–2538

57 Llinas, R. (1992) in Induced Rhythms in the brain (Basar, E. et al., eds),

pp. 269–283, Birkhauser

58 Whittington, M.A. et al. (1995) Synchronized oscillations in

interneuron networks driven by metabotropic glutamate receptor

activation Nature 373, 612–615

59 Wang, X.J. and Buzsaki, G. (1996) Gamma oscillation by synaptic

inhibition in a hippocampal interneuronal network model J. Neurosci.

16, 6402–6413

60 Gray, C.M. and McCormick, D.A. (1996) Chattering cells: superficial

pyramidal neurons contributing to the generation of synchronous

oscillations in the visual cortex Science 274, 109–113

61 Engel, A.K. et al. (1991) Interhemispheric synchronization of oscillatory

neuronal responses in cat visual cortex Science 252, 1177–1179

62 Gray, C.M. and Viana Di Prisco, G. (1997) Stimulus-dependent neuronal

oscillations and local synchronization in striate cortex of the alert cat

J. Neurosci. 17, 3239–3253

63 Young, M.P. et al. (1992) On oscillating neuronal responses in the

visual cortex of the monkey J. Neurophysiol. 67, 1464–1474

64 Tovee, M.J. and Rolls, E.T. (1992) Oscillatory activity is not evident in the

primate temporal visual cortex with static stimuli NeuroReport 3, 369–372

65 Lamme, V.A. and Spekreijse, H. (1998) Neuronal synchrony does not

represent texture segregation Nature 396, 362–366

66 Engel, A.K. et al. (1992) The functional nature of neuronal oscillations

(reply) Trends Neurosci. 15, 387–388

67 Livingstone, M.S. (1996) Oscillatory firing and interneuronal

correlations in squirrel monkey striate cortex J. Neurophysiol. 75,

2467–2485

68 Tovee, M.J. and Rolls, E.T. (1992) The functional nature of neuronal

oscillations Trends Neurosci. 15, 387

69 Kruse, W. and Eckhorn, R. (1996) Inhibition of sustained gamma

oscillations (35–80 Hz) by fast transient responses in cat visual cortex




Now available online… Trends in Cognitive Scienceswww.trends.com

When you access this site you will be asked to complete a simple form (this usually takes aroundtwo minutes to complete) that will allow you access to BioMedNet, the website on which the trendsjournals are located. You will then be allocated a password that will allow you free access to thetrends online web page until May 1st 1999. You should then select the link to Trends in CognitiveSciences to begin searching the journal.

The online version contains a simple search and browse facility that will enable you to look at a fulltext version of all articles published within the journal since January 1998. Browsing and searchingthe articles is straightforward and designed around the need for simple navigation. Articles can beviewed in three display formats: with thumbnail figures, with full figures or as a text-only article(with pop-up figures) for rapid downloading. High-quality printouts can also be obtained from thePDF files. Other features include a full-text archive starting from January 1998 (we plan to extendthis to the first issue April 1997 in the near future), a record of your previous searches, and an e-mail contents alerting service.

I hope you will take the opportunity to access the gateway site at www.trends.com to exploreTrends in Cognitive Sciences and any of the other 15 Trends journals that are temporarily availableto all users at the site until 1 May 1999. From 1 May 1999, only personal print subscribers to Trendsin Cognitive Sciences will benefit from continued online access to Trends in Cognitive Sciencesonline by accessing the site using a unique subscription key. Subscribers and nonsubscribers alikewill still have access to full contents lists and abstracts, search facilities and the e-mail contents alertservice and it is possible to download full-text articles on a pay-per-view basis via BioMedNet.

Finally, I want to provide a service that meets the needs of our readers. I am therefore very inter-ested in your opinion of Trends in Cognitive Sciences online and the benefits it offers. I would begrateful for any feedback concerning the online version, any problems you have encountered, andhow we could improve this service. If you complete the trends online questionnaire available on thecontents pages of all online journals, you will also be entered into a draw for a $500 cash prize!

I hope that you enjoy what Trends in Cognitive Sciences online can offer you now and look forwardto offering you further benefits in the future.

Peter Collins Editor

Documents

Oscillatory gamma activity in humans and its role in object … · ed.), pp. 95–110, Elsevier 38 Aggleton, J.P. and Shaw, C. (1996) Amnesia and recognition memory: a re-analysis